Content
67%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A thorough, highly actionable operating-mode skill with explicit validation in its destructive workflow and well-structured external references. Its main weaknesses are cross-section repetition that bloats the token budget and bulky inline reference material that could be split out.
Suggestions
Consolidate the repeated single-owner / single-path / deletion principles that recur across Greenfield Vision First, Execution Loop, Finding Format, and Final Check into one canonical statement referenced by the others, to reduce redundancy.
Move the Smell Catalog and/or Finding Format template into a reference file linked from the body, leaving SKILL.md as a leaner overview that points to the detailed catalogs on demand.
Make the verify-failure feedback loop explicit in the Execution Loop (e.g., 'if verification fails: fix and re-verify; only when it passes proceed to deletion') to reach the top workflow-clarity anchor.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense original methodology with no basic-concept padding, but ~6 txt blocks restate the same single-owner/single-path/deletion principles across sections (Vision First, Execution Loop, Finding Format, Final Check). Not a 2 because it never explains concepts Claude already knows; not a 4 because the cross-section repetition exceeds 'minor'. | 3 / 5 |
Actionability | Provides concrete, executable guidance — a 7-step context-gathering list citing 'Count callers and imports with rg', a 10-step execution loop with specific file operations, and fill-in templates. Not a 5 because the templates are blank schemas rather than worked examples covering common cases. | 4 / 5 |
Workflow Clarity | The destructive execution loop has explicit validation checkpoints ('Verify with targeted tests, typecheck, and relevant smoke coverage' before 'Delete the old path', then 'Re-run rg for stale names') plus a Final Check checklist and Stop-and-Ask gating, so the destructive-operation cap does not apply. Not a 5 because the verify-failure retry loop is implied rather than spelled out as a feedback loop. | 4 / 5 |
Progressive Disclosure | Two clearly signaled, condition-gated, one-level-deep references to verified-real files (wave-ordering.md, goal-template.md) plus a Related Moves section give good navigation. Not a 5 because sizable reference-style material (Smell Catalog, Finding Format) is kept inline, a minor organization gap. | 4 / 5 |
Total | 15 / 20 Passed |